Risk Adjustment Field Notes · V28 Engine

A RADV auditor never reads your intent. They read your chart.

MEAT is not a coding nicety. In 2026 it is the difference between a diagnosis that holds and one that gets extrapolated across your whole contract. Here is how the V28 engine catches the thin ones first.

TARGET / RADV MEAT documentation 2026 MODEL / CMS-HCC V28 READ / 7 min

The short answer

Submit an HCC that the note does not Monitor, Evaluate, Assess, or Treat, and under CMS RADV extrapolation a single unsupported diagnosis stops being a one-chart problem. It becomes a sampled error that CMS projects across your contract. The fix is not more coders reading charts after the fact. It is a validation layer that flags the missing MEAT before the claim ever ships.

Why 2026 is different

The FFS Adjuster is gone. Extrapolation is in.

CMS finalized its RADV methodology in the January 30, 2023 Final Rule. Two decisions reshaped the risk math for every Medicare Advantage organization, and they are fully in force for the audits landing now.

1No more FFS Adjuster

CMS removed the Fee-for-Service Adjuster that once softened audit findings.
The error rate CMS finds in the sample is the error rate you own.
No built-in cushion for the gap between MA coding and FFS claims data.

2Extrapolation applies

CMS extrapolates RADV findings starting with payment year 2018 audits forward.
A sampled unsupported HCC projects across the enrollee population it represents.
The exposure is the projected recovery, not the dollar value of the one chart.

Reference: CMS-4185-F, "Medicare Advantage Risk Adjustment Data Validation" Final Rule, effective 2023. Extrapolated recovery applies to contract-level RADV audits from PY2018.

The auditor is not asking whether the patient had the condition. They are asking whether your note, for this date of service, proves you managed it.

The MEAT standard, in one sentence

The support an auditor accepts

MEAT is the evidence, not the label

A diagnosis code in an assessment line is a claim. MEAT is the proof behind it. An auditor validating a submitted HCC looks for at least one of these four, tied to the same encounter, in the provider's own documentation.

M Monitor

Signs, symptoms, disease progression or regression tracked over time.

E Evaluate

Test results, medication response, physical exam findings reviewed.

A Assess

Ordering tests, discussion, records review, counseling on the condition.

T Treat

Medications, therapies, referrals, procedures, plan for the condition.

×What gets disallowed

Diagnosis appears only in the problem list, carried forward untouched.
Code pulled from a prior year and never re-documented at this visit.
Assessment names the condition but the note shows no M, E, A, or T.
Unsigned or unauthenticated encounter; no valid provider attestation.

✓What holds up

Condition assessed with a documented status and a management action.
Coded to full ICD-10-CM FY2026 specificity, matching the narrative.
Signed, dated encounter from an acceptable provider type and setting.
MEAT lives in the same DOS the risk-adjusting code was submitted from.

Why one thin note gets expensive

The extrapolation chain

This is the mechanism, not a forecast of your numbers. It shows why an unsupported HCC caught in a sample does not stay contained to that sample.

Step 1 · Sample

Chart pulled

CMS selects enrollee-year records and requests the supporting medical record.

Step 2 · Validate

MEAT check

Each submitted HCC must be substantiated by the note for that date of service.

Step 3 · Error rate

Unsupported

Codes without MEAT become confirmed discrepancies in the sampled group.

Step 4 · Project

Extrapolate

The sampled error rate is applied across the enrollee population it represents.

one weak chart → sampled error → contract-level recovery

Inside the V28 engine

Where the flag fires, before the claim ships

The engine reads the note the way an auditor will. It de-identifies before any model call, maps the narrative to V28 HCCs, and holds any code it cannot back with same-encounter MEAT.

De-identify at the boundary

PHI is stripped fail-closed before the chart reaches any language model. No PHI egress, ever.

Extract every documented condition

The note is parsed for stated diagnoses and their management language, not just the assessment line.

Map to CMS-HCC V28 and ICD-10-CM FY2026

Each condition is coded to full FY2026 specificity and mapped to its V28 risk-adjusting category.

Run the MEAT test on each HCC

For every risk-adjusting code, the engine looks for at least one of M / E / A / T tied to the same date of service.

Flag, do not submit

Any HCC without same-encounter MEAT is held for review and routed back to the provider, not billed on faith.

Engine hold · MEAT not found

EncounterDOS 2026-03-14 · established patient

Submitted HCCChronic condition mapped to a V28 category

MEAT foundNone · problem-list carry-forward only

RADV riskUnsupported · extrapolation-eligible

Engine actionHold → query provider for status + plan

Archetype shown for illustration. No real patient, provider, or client record is depicted. The engine surfaces empty states rather than filling gaps with assumed documentation.

The 2026 rules this rests on

Cited by name, so you can check us

CMS-HCC Risk Adjustment Model V28 The current CMS risk model, phased in and used for MA risk score calculation. V28 is the mapping target for every code the engine validates. CMS Risk Adjustment

CMS RADV Final Rule (CMS-4185-F, 2023) Removed the FFS Adjuster and confirmed extrapolation of contract-level RADV audit findings from payment year 2018 forward. Audit Methodology

ICD-10-CM Official Guidelines, FY2026 Effective October 1, 2025. Governs code specificity and the requirement that documentation support the code selected for each encounter. Coding Specificity

MEAT documentation standard The Monitor, Evaluate, Assess, Treat framework auditors use to confirm a chronic condition was actively managed at the encounter it was coded from. Support Criteria

See which of your submitted HCCs an auditor would disallow today.

We will run a MEAT-gap read on a de-identified sample of your risk-adjusting charts and hand you the flagged list, before a RADV sample does it for you and extrapolates the bill.

Request a MEAT-gap read

ASP-RCM Solutions · Senior Partner, Frisco · Risk Adjustment & Coding Integrity

Educational content for revenue-cycle and compliance teams. Not legal, coding, or actuarial advice. Regulatory references are cited to the source publications above; verify current CMS guidance and ICD-10-CM FY2026 files against the official releases before acting. No patient, provider, or client data is represented on this page.